no. 90-003...1. introduction 1.1 vsat architecture a communication satellite can be used to provide...
TRANSCRIPT
UNIVERSITY OF HAWAII LIBRARY
~NIVrFISITY OF HAW,\ ?! l.l!"::l,",il'{
PRilSM WORKING PAPER
No. 90-003
VSAT DATA NETWORK
NORMAN ABRAMSON
ABSTRACT
VSAT DATA NE'l'WORXS
Norman Abramson Department of Electrical Engineering
University of Hawaii Honolulu, HI 96822
Two general methods have been used to provide random access packet communications in Very Small Aperture (VSAT) data networks -- Spread Spectrum (or CDMA) and ALOHA. In this paper we review the use of satellite channels for such networks and discuss certain basic aspects of the architecture of VSAT data networks. Although Spread Spectrum and ALOHA have different origins and are sometimes r epresent ed as competing technologies, they can in fact be characterized as different ways of viewing the same low dimensional signals in a high dimensional s ignal space . After a brief introduction to the architecture of VSAT networks we show how a simple linear transformation of conventional ALOHA packets leads to signals identical in all r espects to the most common form of Spread Spectrum signals. We call the result of this transformation spread ALOHA.
There are two practical consequences of this theoretical result. First, for the case of small earth stations it is not possible to find an access technique with a higher throughput than Spread ALOHA. Second, the . use of different spreading sequences for different users in a packet network using Spread Spectrum is not necessary for user separation.
l..
VSAT DATA NETWORKS
Table of Contents
INTRODUCTION
1.1 1.2 1.3 1.4
VSAT Architecture Access Techniques History .... ..... Satellite Multiple Access
.. ....
2. NOTATION
2.1 2.2
Packets and Matched Spreading Sequences
3. SPREAD ALOHA
Filters .......
3.1 Spreading the Packets . . . . . . . . . . . . . . . ..... 3.2 B,it Spread ALOHA . . . . . . . . . . .. 3.3 Discussion .. . . . . 3.4 Conclusions . . . . . . . . . . . . .... . . . .
i
Page
l 4 8
10
14 18
. . . . 19 • • • 20
26 27
1. INTRODUCTION
1.1 VSAT Architecture
A communication satellite can be used to provide conventional
point-to-point channels for data networks and other applications.
But the key characteristic of communication satellites for many
applications is the broadcast and multiple access nature of the
satellite channel. For the physics of the satellite path
provides the built in capability of one to many (broadcast)
channels and of many to one (multiple access} channels to and
from any earth station in the satellite footprint. It is this
characteristic of satellite communications which can be used to
provide new types of service which are simply not practical or
not economical using conventional ground communications.
The broadcast nature of the satellite channel encouraged the
development of satellite communications for video and audio
distribution in the l970's. This same broadcasting capability of
satellite channels bas been employed for a variety of one way
data distribution networks since the early 1980's. During the
last few years we have seen the development of a powerful and
flexible network architecture which coll!hines the capabilities of
satellite broadcasting and satellite multiple access for two way
data networks composed of very small aperture terminals (VSAT's).
The network architecture used in VSAT data networks is almost
always designed around a single large (hub) earth station
transmitting data in a broadcast channel to a large number of
VSAT's.
l
-i-i ... -i Hub USAT'a
Figure 1: A VSAT Network
The VSAT's in such a network transmit data in packets to the hub
station using the multiple access capability of the satellite
channel. Since there are no direct links from one VSAT to
another, any VSAT to VSAT traffic must take a two hop path from
its source to its destination. There are two good reasons for
such a limitation. First, transmit power and r eceive power
requirements on the l arge nwnber of VSAT's in such a network can
be relaxed considerably by the fact that the uplink and downlink
in the network each benefit from the higher performance
capabilities of the hub station. Second, many of the
applications which can be nefit from the use of VSAT t echnology
require communications from a large number of users to a central
information resource. These appli cations include large financial
networks, reservation networks and shared data bases.
2
The link from the hub station of a VSAT network to ~he VSAT's is
easily confi gured using conventional time di vision multiplexing
(TOM), And although there are differences in data rates, in
modulation techniques and in transmission formats among different
VSAT networks there i s general agreement on the use of TOM for
multiplexing from the hub to the terminals.
µ
----~ ~ ii · ... i
USAf0
•
Figura 2: Broadcast Channel - Hub to VSAT's
The multiple access link from the VSAT's to the hub however, has
been subjected to a greater degree of variation in data networks
which have been built during the past decade. And in the overall
design of the network the choice of .an access technique for the
VSAT to hub multiple access channel is the key decision which
must be made by the network architect. There are still a nwnber
of important choices in data rates, modulation t echniques and
transmissi~n formats. But while these differences will have a
maJor impact on overall network performance i t i s probably
accurate to characterize the ·choice of access technique as the
primary feature which distinguishes one network from anothe r. In
this paper we focus on the choice of access techniques for VSAT
networks and analyze the architecture of these networks,
3
; ):( .. ;
; ;
' ; ,:.
It
....
Figure 3: Multiple Access Channel - VSAT•s to Rub
1.2 Access Techniques
Frequency Division Multiple Access (FDMA) and Time Division Multiple
Access (TOMA) are the two primary access techniques used for sharing
the high ~pacity of a typical commercial satellite transponder
all!ong several earth stations in a voice network. With either of
these t echniques the total transponder capacity is divided into
frequency or time slices for use in fixed allocation channels of
moderate capacity. When the traffic from a single tet'lllinal in a
network is bursty the efficiency of this kind of operation drops
and a De:mand Assigned Multiple Access (DAMA) t echnique is
required in order to allocate capacity in response to fluctuating
user demands.
In a DAMA system network capacity is allocated (either in
frequency or time slices) in res ponse to user requests for
channel capacity. Note however that in one sense the use of a
DAMA system merely shifts. the problem of allocating capacity in
response to random requests to another l evel. A special channel,
usually called the request channe l, is assigned to carry requests
for capacity assignments in a DAMA network and some access method
4
to allow reasonably efficient sharing of the request channel mus t
be found.
During the past few years it has become practical to build VSAT
networks composed of hundreds or thousands of small (typically
1.2 meters in di!lllleter) t erminals. The traffic in these networks
is usually in the form of single data packets originating from
interactive users or bursts of data packets originating from some
file transfer protocol. In general as the number of earth
stations in a network increases the more the traffic from a
single station will appear to fluctuate due to random user
demand. In such a network of course the use of fixed assignment
FOMA or TOMA is usually impractical, while the use of DAMA would
ordinarily impose an unreasonable amount of overhead in the
network. In order ~o provide multiple access to packet data
networks, therefore two other access techniques have been used
--- Spread Spectrum Multiple Access (or Code Division Multiple
Access · - COMA) (l] and ALOHA (2) (in a so called unslotted or
slotted mode).
Spread spectrum has been defined as communications which uses a
"bandwidth well beyond what is required to transmit digital data"
(3). ln Direct Sequence Spread Spectrum, the most common form of
spread spectrum, each bit to be transmitted is converted into r
binary chips, where r is much g~eater than one and the original
signal bandwidth is expanded by the same factor of r in order
to accornmodate the higher transmission rate of the chips. Each
transmitter in such a network is usually assigned a distinct
5
spreading sequence and the receiver must then convert the stream
of .chips from each transmitter into a (lower data rate) stream of
bits by means of what is essentiall y a digital matched filter,
one matched to each distinct transmitter spreadi ng sequence. A
frequency hopping version of this procedure is also possible, but
frequency hopped spread spectrum seems better suited to military
rather than commercial applications.
packet
chips
Figure 4: spread spectrum -- From Bits to Chips
In the .1µ.0HA form of multiple access the bandwidth of the channel
is also required to be greater than the infonnation bandwidth
corresponding to the average data rate transmitted from any
single user in the channel. Each transmitter in such a multiple
access channel will transmit its data packets at the maximum data
rate of the channel. Because the channel bandwidth is hi gh
compared to the data rate from a single user, the duty cycle of
each user is low and the probability that two packets from
different users will overlap in the channel will be low. A
variety of protocols have been developed [4,5) to deal with the
case of packets which are lost due to contention in the channel.
6
The ratio of the maximum data rate which can be sustained in such
a channel to the maximum data rate when the channel is used by a
single user in a point-to-point mode is called the maximum
throughput of the ALOHA channel, and is equal to l/2e or o.184
(2). If the channel is modified so that each transmitter
synchronizes the beginning of its packet transmissions to fixed
slots then the maximum throughput of the channel is doubled to
l/e or 0.368 (6) .
t -
Figure 5: ALOHA Multiple Access
Note however that both the l/2e and the 1/e maxi.mum throughput
results compare the throughput of a low duty cycle (typically 10%
or less) ALOHA channel to that of a point- to-point channel
operated continuously. The continuous channel of course uses
considerabluy more power (typically 10 db. or more) than the
bursty ALOHA channel. For the case of the original packet radio
ALOHANET (2) whers the~ power is important such a throughput
comparison is appropriate. However in the case of most satellite
multi-access channels, and certainly in the case of VSAT
7
channels, where the limiting resource is the average power
transmitted in the do'W?!link, a different kind of comparison is
appropriate. In this case it makes more sense to compare the
data rate of the ALOHA channel with the data rate of a point-to
point channel using the same average downlink power as the ALOHA
channel. And as might be expected, when proper allowances are
made for a 10 db. or more power mismatch, the ALOHA channel
throughput looks considerably more attractive. In s~ction 1.4 we
discuss the throughput os an ALOHA channel with an average power
rather than a peak power constraint.
1,3 Jlistory
Each of the two multiple access techniques described above was
originally developed for applications other than the satellite
channel. Spread Spectrum has its origins in a variety of
military applications [7) and much of the early work in spread
spectrum dealt with the transmission of continuous signals rather
than the packet transmissions of interest in most present day
systems. ALOHA channels were originally developed in the early
1970's for use in the ALOHANET, a Local Area Wireless Network
{LAWN) at the University of Hawaii [8]. The first demonstration
of the use of either of these techniques for satellite
communications however was in 1973 in the ALOHA SYSTEM'S PACNET
network, using the NASA ATS-1 satellite to link VSAT's in Hawaii,
Alaska, Japan and Australia .
8
In 1976 an unslotted ALOHA channel was implemented as the request
channel of the COMSAT MARISAT s ystem [9) in a commercial network
--- the first commercial application of this technology. The
first commercial use of spread spectrum in a satellite channel
was in 1981 with the introduction of the Equatorial
Co!lllllunications "C-100 micro earth station" for satellite data
distribution . The MARISAT ALOHA channel was a true multiple
access channel, providing many to one communications to thousands
of small ship stations in a three ocean system. The first
Equatorial Communications terminals were receive only units in
which only the interference rejection capabilities of spread
spectrum signals were utilized for data transmission. In 1984
Equatorial introduced a true multiple access satellite network
using the "C-200 Micro Earth Station". By 1985 several other
c~mpanies had entered the multiple access VSAT market in the US.
Ironically, by 1987, Equatorial which had first demonstrated the
commercial viability of the VSAT concept had run into troubl e in
its c-200 systems, and had been taken over by contel . By 1986
this market had matured so ·that several other companies had
entered the market and by the end of 1989 there were four major
supplier~ of this equipment in the USA -- TRIDOM/AT&T, Hughes
Network Systems/GM, GTE/Spacenet and Equatorial/ConTel. Each of
t~~~e systems ~as based on an ALOHA channel (either slotted or
unslotted) in some cases supplemented by TOMA for large file
transfers, although ConTel also offered its original Spread
Spectrum Multiple Access network as an option. By 1989 there
were about 10,000 multiple access commercial VSAT terminals in
9
operation not including the large number of ship stations using
ALOHA in the INMARSAT (formerly MARISAT) system.
1.4 satellite Multiple Access
Although both Spread Spectrum and ALOHA can provide a multiple
access capability in many different kinds of c9mmunicati9n
networks, there are a number of special characteristics of the
VSAT packet data multiple access channel which affect the
operation of the overall system. When spread spectrum is used t
distinguish among large numbers of intermittent packet
transmitters each using a different spreading sequence, the hub
receiving station must be provided with a digital matched filter
operating at the high speed chip rate for each of the possible
transmitters. In the Equatorial c-200 system these receivers ar
called "ear cards". And in a network with thousands of VSAT's
the logistics of installing, removing, maintaining and testing a
special piece of hardware at the hub station for each network
user is a difficult task.
The throughput of an ALOHA channel is ordinarily calculated on
the basis of comparing the average data rate of the burst ALOHA
channel to that of a full time dedicated point to point channel
between only two users. In the case of ground communications
where the average power radiated by the transmitter is not a
major constraint such a comparison is appropriate. But the
bursty ALOHA channel (see Figure 5) will ordinarily require 10%
or less of the average power required by the point to point
channel in such a comparison. In the case of a satellite channe
10
the average power transmitted from the satellite transponder is a
limited and expensive resource, so that a different kind of
comparison is appropriate.
For the satellite channel it is appropriate to begin the
comparison using Shannon's basic equation for the capacity of th~
bandlimited additive white Gaussian noise channel. In this case
R, the data rate of the channel in bits per second, is limited by
the channel capacity
(1)
where the logarithm is taken to the base 2, w is the channel . '
bandwidth, Sis the average signal power and N is the average
noise power. Shannon's equation assumes that the signal is
continuous, but if the time bandwidth product of the packets is
large enough we can apply Shannon's equation to the packets
during the tillles of packet transmission. And if the duty cycle
of packet transmissions is d { d << l) the data rate of the
multi-access ALOHA channel must be less than
C = dC 1 O
s = d w tog ( 1 + N) (2)
It is not difficult to find the corresponding capacity of a
point-to-point channel with the same average power as the ALOHA
channel . For a given duty cycle, d, the average power required
by the ALOHA channel is just dS, and the data rate of the
continuous channel must the r efore be less than
ll
(3)
Because of possible packet overlap equation (3) is valid for d<<l
and certain corrections must be made for the more general case.
These calculations are provided in [lO ; section VJ where the exact
derivation is provided for four separate cases the slotted and
unslotted ALOHA channel, operating with either a linear or a hard
limited transponder . In all cases the the basic results are the
same. As might be expected we always have
(4)
That is, the multi-access ALOHA channel will always have a lower
maximum throughput than the point-to-point channel with the same
average power. However, for all values of the duty cycle, the
ratio of the maximU111 ALOHA throughput compared to the capacity of
the average power limited channel exceeds the ratio of the
maximum ALOHA throughput compared to the capacity of the peak
power limited channel by a significant amount. And for the case
of a low signal-to-noise ratio in the receiving antenna (e.g.
VSAT's)
(5)
12
Thus we have the surprising result that f or the cas~ of small
s atellite earth stations operating with a low duty cycle, the
ALOHA capacity approaches the Shannon capacity, and it is not
possible to find an access method with a higher data throughput
[10; section VJ.
For the kind of ALOHA operation described in the previ ous
paragraph, in order to approach the Shannon channel capacity it
is necessary to operate at a low duty cycle, and therefore for
interesting values of average power, the peak power of the VSAT
trans~itter during data bursts can be quite high. And this
characteristic of the low duty cycle ALOHA channel could limit
the application of this mode of operation in satellite networks
composed of large numbers of VSAT's . Since the signal
detectability depends upon the energy per bit and not the average
signal power it makes sense to spread the transmitted packets in
time in order to decrease the peak power requirements of the
earth station transmitters while keeping the transmitted energy
per bit constant. Note that this strategy can be used in the
case of a burst ALOHA channel without decreasing the average data
rate of the channel . In section 3, we show how it i s possible to
spread the packets of a low duty cycle ALOHA channel in time,
without increasing the probability of packet overlap out of the
channel receiver. But first we use section 2 to i ntroduce a
notation which will be helpful in the remainder of this paper.
13
2, NO'rATION
2,1 Packets &lid Matched Filters
Consider the case of packet transmission where each packet consists
of exactly n bits. Let the value of these bits be dj, where the
data ·sequence dj is given by
~-±1 j-0,1 ... n·l (6)
In this paper we s hall be concerned with binary phase-shift keyed
(BPSK) channel s for transmission of packets and with the
representation of these packets in a form which emphasizes their
correspondence to conventional direct sequence spread spectrum
channe l s [11). Let p(t) be the bit pulse waveform used in the
channel so that a single packet, D(t), consisting of n bit
transmissions at the times 0,1, • • • n-1
n·l
O(c) • L dJp(C•j) j-0
can be represented as
(7)
A useful device which allows us to separate the analog effects of
pulse shape from the digital effects of the data sequence is to
represent Q(t) as the co~volution of the pulse p(t) wi th a sequence
of impulses multiplied by the data sequence values dj. Then
•·l D(t) • L d,&(t·j) * p(c)
J•O (8)
Note that D(t) can be thought of as the output of a time invariant
linear operator with impulse response p(t), when the input is
14
n•l
d(e> - I dj6<t·j> (9) J••
d(t) D(t) I p(t) I Figure 6: The Packet as the output of a Linear Filter
The reception of a sequence of packets occurring at random times
in a multiple access channel is usually broken down into two
stages a packet detection and synchronization stage (to
detect the presence of the packet and to synchronize) and a
signal detection stage (to demodulate the signal and to determine
the values of the information bearing elements. In a VSAT
network the synchronization problem is made considerably simpler
by the fact that the hub station can provide information feedback
to the VSAT terminals in order to help them synchronize the phase
of their packet transmissions. We therefore need only consider
the signal detection part of the receiver. We assume the packet
O(t) is transmitted in a channel wi th additive white Gaussian
noise (AWGN), n(t), so that the received signal a(t) is
a(t) • D(t) + n(t) (10)
15
Under a wide variety of assumptions, a key element of the packet
receiver [12,13] is the matched filter matched to the pulse
shape, p(t), Ignoring questions of time delays in the receiver,
we can represent the impulse response function of the matched
filter as p(-t), and the output of the matched filter as b{t)
[ 14, 15] •
a{t) • D(t) + n(t)
I b(t) p(-t) •~~~-
Figure 7: Input and output of the Matched Filter
Then the output of the matched filter is given by
li(t) • a(t) • p(·t)
- D(t) • p(•t) + n(c) • p(•t)
n•l
•}: dJ6(t·j) • p(c) • p(-c) + n(c) • p( -c) J•Q
Define the correlation of the bit waveform as
p(C) • p(C) • p( • C)
• • f p(X)p(C•X) d.~ ...
Then,
16
(11)
(12)
•·1 b(t) • 1 dj6(t•j) • p(t) + n(t) • p(·t)
J•O
n•I (13) - r djp(t•j) + n,(t) J•O
where we define
n1(t) • n(t) * p(·t) (14)
Many decision rules to obtain the data sequence, dk, from the
received signal rely on sample values of b(t) taken at t equal to
the bit transmission times k = O,l,2 •••
o•I
b(k> - E djp<k·j> + n 1<k> J•O
• <1i,p(O) + I(k) + n1(k) (15)
where n•l
I ( k) - L djp(k·j) j-0
J•k ( 16)
is the inter-symbol interference for the k'th symbol in the packet.
The energy per bit is
• 2
~ - f p (t) dt ...
Then, if N0 is the power spectral density of the noise, E!)IN0
provides a measure of the quality of the signal detection phase
of the receiver [16).
17
(17)
2.2 Spreading Sequences
In order to achieve the time spreading of the packets discussed
in section 1.4 we use binary sequences with low autocorrelation
properties, such as Barker sequences [17,18]. Define a binary
sequence of length r as
j • O,l ... r-1 ( 18)
and a time invariant linear operator (spreader) corresponding to
the sequence sj with an impulse response function
.c-1 .
s(t) • L sJ6(t·j) J•O
Define the correlation of s(t) as
where
a(t) • s(t) • s(-t)
• - J s(x~ s(x+t) dx -
r-1
• l "J6(t·j) J•l·r
r·l
"J • sJ•sJ - l sk s,. J k•O
18
(19)
(20)
(21)
In the case of Barker sequences,
(22) for j•O
and
for J"'O
(23)
Barker sequences are known only for the cases of r-2,3,4,5,7,ll
and 13, but binary sequences of other lengths with good values of
for use in Spread ALOHA channels as well as efficient methods of
searching for such sequences have been obtained by Li (18).
3, SPREAD ALOHA
3,1 Spreading the Packets
As explained in section 1.4, the low duty cycle ALOHA channel shown
in Figure 1.5 can be used in VSAT networks with one major theoretica
advantage (it can achieve the Shannon bound) and one major practical
disadvantage (it may r equire a high value of peak power out of the
small aperture earth station). In this section we modify the
signals in the low duty cycle ALOHA channel so that the disadvantage
of high peak power is eliminated while at the same time EbJ'N0 the
measure of signal quality defined in section 2.l remain unchanged.
We call this modification "Spread ALOHA", and our results show that
the performance of Spread ALOHA is the same as that of low duty
cycle ALOHA while the performance of the latter is the best possible
for a given average power and a given channel bandwidth.
19
The modification we use is a natural one --- we take the packets
shown in Figure 5 and spread them in time. such time spreading can
be done in a number of ways, but the method we describe in section
3.2 has the property that the spreading leaves Et)IN0 unchanged and
at the same time the spreading does not affect the probability of
packet overlap at the output of the detector. If the satellite
channel is operated in a linear mode, the increased channel overlap
will not affect the operation of the ALOHA channel since the optimum
detector consists of a matched filter which serves to despread the
spread packets .
3.2 Bit Spread ALOHA
our objective then is simply to spread the in time, thus effectively
lowering the transmit power requirements of the VSAT terminals,
while at the same ' time keeping the transmitted energy per bit
constant, and also by means of the despreading operation at the
network hub, keeping the effective overlap silhouette of the
transmitted packets constant. Note the distinction made here
between the overlap of packets in the physical channel which is
greatly increased by Spread ALOHA, and the overlap of packets at the
output of the detector which is not.
For Bit Spread ALOHA packets we start with equation (7) and delay
the transmission of each bit by r units of time relative to the
previous bit. That is, let the transmitted packet be
20
n•l
D,(t) - 1 dJp(t-rj) J•O
and, following the development of equations (8) and (9)
n· l
· D,(t) - I dJ6(t -rj) * p(t) J-0
(24)
(25)
As before we can represent Or(t) as the output of a time invariant
linear operator with impulse response p(t) , when the input is
•·1 d,(t) - 1 dJ6(t-rj)
J-0
p(t) I
(26)
D,(t)
Figure a: The stretched Packet as the output of a Linear Filter
We refer to the packet Or(t) as the stretched packet. And we
produce the spread packet we want by passing the stretched packet
through a spreader, a time invariant linear filter with impulse
response equal to r-l/2 s(t), as given in equation (19).
p(l:) I D,(t)
I D
1(t)
,:·112 s(t: •-----
Figure 9: The Bit Spread Packet as the output of Two Linear Filters
21
The process of spreading an ALOHA packet for a simplified packet is
illustrated i n-Figure 10. In order to clarify the process the bits
and chips are represented without the pulse shaping filter p(t).
The first line of Figure lO depicts a packet composed of 6 bits --
llOlOl . The stretched packet of equations (24) and (26) is shown on
the second line. The third line illustrates the spread packet for
this simple example, where the spreader is taken as the length 7
binary sequence of line 4 --- 1110010 . Of course more realistic
values of both the packet length and the length of the spreading
sequence would be in the range from lOO to 10,000.
II I I Packet : 118181
I I
_.j._ __ -'---~----11--------'--- Stretched PacJcet I
' I I I t I
II! I Ill I ; II IIIJ I : IJ llfl J Spread Packet
I I I I I I I I I I I I I I I I I I I I
l I I I Spre.acl i 119 Se<{uenc.e: 1110818
II I Figure 10: EXa.mple of a spread ALOHA Packet
The form of the spread packet is now easily calculated as
D,(t) - r·112 d,( t) * s( t) * p(t)
rn-1
- t cjscc-J> * p<c> j-0
rn- 1
- t cjp<c·J> j-0
22
(27)
where
C - r·l/2 d s. j O ~
(28)
and «: is the largest integer less than or equal to j/r and /3 is equal to j modulo r
a - LJ/rJ ,8 • j mod r
(29a)
(29b)
The binary random variables, cj, constitute the "chips" of the
Spread ALOHA signal; more precisely, cj is the /3 'th chip of the
"<' 'th data bit of the spread signal. Since each bit of the original
packet is converted to r chips of the spread packet by the
spreader, each chip is multiplied by r-1/ 2 in order to maintain
the same energy per bit before and after spreading. Thus the peak
power requirement of the transmitter is decreased by a factor of r .
Time spreading provides a mechanism for reducing the average power
of the transmitter to a level consistent with a network of VSAT
terminals while maintai ning a constant value for the energy per bit.
Note that in contrast to conventional spread spectrum for continuous
signals, each packet in the Spread ALOHA channel is spread by the
same binary sequence, s j . Separation of packets from different
users is accomplished by means of the ALOHA contention protocol
rather than by the cross-correlation of different spreading
sequences. And in contrast to conventional operation of an ALOHA
channel the separation is .not limited by the ove rlap of the
transmitted packets in the channel, but by the overlap of these
pulses at the output of a matched filter .
23
From Figure 9 we see that the problem of detection of the bits of
the spread pa~ket in the presence of AWGN has been reduced to that
of the unspread packet of Figure 2. We need only replace p(t) in
Figure 6 by the convolution of p(t) and s(t) in Figure 9. Then
f or the spread case the spread matched filter is shown in Figure ll.
b(t) a(t) • 01 (t) + n(t) I I I -------:__p•(•-•t•)-• r"
112
s(-t: •-----
Figure 11: Input and output of the Bit spread Matched Filter
Now following the development used in equation (ll} we have
b(t) - r"112[a(t) • p(•t) • s(·t)]
- r·112 [D (c) • p(·t) • s(·t)] + r"112 [n(t) • p(·t) • s( -c:)] • - r"1{d (t) • s(t) • s( •C:) • p(c) • p( ·t) ] + r·112[n(t) • p( ·t) • s(·t)]
<
where we define
llz(C) - r"112(n(C) • p(•t) • S(•C))
(30)
(31)
Using (20) and (26), we can expand b(t), the output of the matched
filter in equation (30)
24
n·l r--1
b(C) - r"1 r dJ"(C•rj) * r c,J6(t•j) * p(C) + tiz(t) J•O J•l•r (32)
If we define d_1 a dn = 0 and <5"" • o , then -r
nt•l
b(c) - I: ~,6(t•j) • p(t) + nz(t ) (33) J•l·t
nr• l - r ejp(c-j) + nz(t) j•l·t
where
(34)
and OC. is the largest integer less than or equal to j/r and /3 i s equal to j modulo r.
a - l)/rj
{J • j mod r
(3!5a)
(35b)
The output of the matched filter is s ampled at the stretched bit
transmission times t•kr, where k~o,1,2 ••• n-1. At these sample
points /3: 0 and a;= r , so that ekr = dk, and we get
25
where
nt·l
b(kr) - L eJp(kr·j) + n2(kr)
J•l·~
• di.P( O) + J(kr) + n2(kr )
.... 1
J (kr) - L eJp(kr- j) J•l·~ J,ikz
is the inter-symbol interference for the k'th symbol in the
packet.
3.3 I>iscussion
(3')
(37)
Note the similarity of the interference term of equation (37) to
that of equation (16) • . The sample values of the data pulse
correlation, ('(t), caus~ the interference. In this case the
interference might properly be termed "inter-chi p interference"
rather than inter- symbol interference since the stretched packet
has separated each bit by r units of time and, for reasonably
shaped pulses, p(t) , this will eliminate the possibility of
inter-symbol interference. And even in those cases where the
sample values of . j'(t) are not negligible note that the
coefficients ej as given in equation (34) decrease as r-1 • This
means that ·the interference caused by adjacent chips in Bit
Spread ALOHA will be considerably less than that caused by
adjacent bits in unspread packets.
26
Finally we note that packet samples utilized by a matched filter
detector of Data Spread ALOHA packets are taken at intervals of r
units of time. In the case of the AWGN channel assumed, packets
which overlap in the channel can still be detected correctly as
long as all samples of each of the two packets are separated by a
single chip interval, taken as a unit of time . Thus the total
period of vulnerability in this unslotted Spread ALOHA channel is
twice the packet length, just as in a conventional unslotted
ALOHA channel. The only difference in this case is that the
period of vulnerability is broken up into 2n-l small
subintervals. In contrast to conventional ALOHA channels two
packets which overlap in the channel do not necessarily . r esult in
lost packets due to collision. The spreading and despreading
operations effectively separate most of the channel collisions in
Spread ALOHA signals so that the only overlap which· is of concern is
the overlap at the output of the matched filter detector .
The spreading technique we have described is not the only
possible method of achieving the practical advantages o! Spread
ALOHA. In (19] a similar dual spreading procedure, called Chip
Spread ALOHA is analyzed and an algebraic coding formulation of
Spread ALOHA is provided.
3.4 conclusions
The arguments o! equations (1) to (5} show that the usual analysis
of ALOHA throughput for peak power limited channels is not
appropriate for the case of average power limited VSAT channels.
The well known l/2e • 0.184 and 1/e = 0.368 limits compare the
27
performance of a multiple access ALOHA channel to that of a point
to-point channel with significantly greater average power in the
downlink. Simply put, some of the results originally published by
this author are too often used as the right answer to the wrong
question.
When the constraint of identical average power in the satellite
downlink is taken into account, the maximum value of the throughput
of conventional slotted or unslotted ALOHA channels will always be
higher than the l/2e or 1/e limits, and in the case of VSAT networks
typical of those available today, significantly higher. The
greatest improvement will occur for the case of low signal-to-noise
ratios in the VSAT receiver and low duty cycle transmission --
precisely those conditions of greatest interest in VSAT networks
(101 section VJ.
For VSAT networks with an average power limitation in the downlink,
as the bandwidth used for the multi-access channel is increased the
signal-to-noise ratio in the VSAT receivers and the duty cycle
defined in terms of the multi-access bit rate (not the chip rate)
both decrease. In the limit the maximum value of the throughput
approaches the Shannon bound and it is not possible to find an
access technique with a higher throughput.
In order to achieve reasonable values of signal energy per bit using
conventional ALOHA operating at a low duty cycle in the high
bandwidth channel it might be necessary to require unreasonably high
values of transmitter power from the VSAT terminals. The solution
to this problem is to spread the VSAT packets in time to obtain a
28
low power Spread ALOHA signal. In this case the Spread ALOHA signal
becomes identical in form to a COMA spread spectrum signal, and in
the limit both types of tranSl!lission achieve the Shannon bound.
Separation of the signals from different transmitters at the huh
station however can utilize the ordinary signal separation mechanism
of the ALOHA channel rather than the chip code separation of the
CDMA channel.
29
REFERENCES
l. Edwin B. Parker, "Micro Earth Stations as Personal computer Accessories", Proceedings of the IEEE, vol. 72, no. 11, November 1984, pp. 1526-1531.
2. N. Abramson, "Packet switching with Satellites," AFIPS Conference Proceedings of the National Computer Conference , New York. Vol. 42, pp 695-702. June 1973
3. Andrew J . Viterbi, "Spread Spectrum communications and Real ities," IEEE Communications Magazine, Vol . 17, 1979, pp. 11-18.
Myths no. 3, May
4 . Dimitri Bertsekas and Robert Gallage r, "Data Networks", PrenticeHall, Inc., Englewood Cliffs, New J e r sey
5. Simons. Lam, "Satellite Multiaccess Schemes for Data Traffic", Proceedings of the 1977 International Conference on Communications, Vol. III, pp 37.1.19-37.1.24, June 12-15, 1977.
6 . Roberts, Lawrence G., "ALOHA Packet System With and Without Slots and Capture", Computer Communication Review, Vol. 5, No. 2; April 1975; pp. 28-42
7. Robert A. Scholtz, "The Spread Spectrum Concept," IEEE Transactic Communications, Vol. COM-25, no. 8, August 1977, pp. 748-755.
8. Norman Abramson, "Development of the ALOHANET", IEEE Transactions on Information Theory, Vol. IT-31, No. 2, March 1985
9. D. W. Lipke, D.W. Swearingen, J.F. Parker, E.E. Steinbr echer, T.O. Calvit and H. Dodel, "MARISAT - a Maritime Satellite Communication System", COMSAT Technical Review, Vol. 7, No·. 2, Fall 1977.
10. N. Abramson, "The Throughput of Packet Broadcasting Channels," IEEE Transactions on Communications, Vol. COM-25, no. 1, January 1977, pp. 117-128.
11. Raymond L. Pickholtz, Donald L. Schilling and Laurence B. Milstein, "Theory of Spread-Spectrum Communications --- A Tutorial," IEEE Transactions on Communications, Vol. COM-30, No. 5, May 1982, pp. 855-884
12. Michael B. Pursley, Dilip V. Sarawate and Wayne E. Stark, "Error Probabili ty for Direct-Sequence Spread-Spectrum MultipleAccess Communications ---Part I: Opper and Lower Bounds ," IEEE Transactions on Communications, Vol. COM-30, N. 5, May 1982, pp. 975-984.
13. Norman F. Krasner, "Optimal Detection of Digitally Modulated Signal," IEEE Transactions on Communications, Vol. COM-JO, no. 5, May 1982, pp. 885-895.
ref-1
14, George L. Turin, "An Introduction to Matched Filters," IRE Transactions on Information Theory, Vol. IT-6, June 1960, pp. 311-329.
15. George L. Turin, "An Introduction to Digital Matched Filters,u Proceedings of the IEEE, Vol. 64, No. 7, July 1976, pp. 1092-1111.
16. George R. Cooper and Clare D. McGillem, Modern Communications and Spread Spectrum, McGraw-Hill Book Company, New York, 1986.
17. R.H. Barker, "Group Synchronizing of Bi nary Digital Syste111S," in Communication Theory, edited by w. Jackson, Academic Press, 1953, pp. 273-287.
18. Ping-Fai Li, "Binary Sequences with Low Autocorrelation," ALOHA SYSTEM Technical Report B86-l, University of Hawaii, Honolulu, May 1986.
19. Norman Abramson, "Spread ALOHA for VSAT's", ALOHA System Technical Report B86-4, University of Hawaii, Honolulu. June, 1986.
ref-2